This study demonstrates the feasibility of incorporating a network of health workers affiliated with community-based organizations into strategies for molecular surveillance of malaria resistance in remote and unstable areas of Myanmar, including priority Tier 1 and Tier 2 areas at high risk for the spread of delayed parasite clearance. This surveillance network can accelerate the assessment of artemisinin resistance and delayed parasite clearance once genetic markers are identified. Myanmar’s ecologic, ethnic and political diversity, coupled with patchwork access to malaria control services and quality anti-malarials creates a mosaic of selective drug pressure, and this study provides a preliminary glimpse of patterns emerging over time and place in remote areas of the country, including areas directly across the border from western Thailand where delayed parasite clearance was recently confirmed. The prevalence of pfmdr1 amplification was relatively low (12-19%) across the three border regions included in the current study; and only a single isolate had a CN of three or more. Results from Karen State contrast with recent studies documenting elevated CN in at least 40% of isolates
 with a mean CN of 2.9 (SE 1.4)
 among refugee, migrant and cross-border populations in Thailand (Figure
1). Findings from Kachin State contrast with a recent study in Laiza that found no pfmdr1 amplification in 171 P. falciparum isolates
[40, 41], but is consistent with a study conducted in 2004 in nearby counties of Yunnan province
. The systematic review failed to identify previous studies of pfmdr1 CN along the borders with India and Bangladesh to compare to the 11% prevalence estimated for that region.
Multiple factors likely contribute to the observed variation in pfmdr1 amplification documented within and between populations living in Myanmar and across its borders in neighbouring countries. Little is known about host response in populations in Myanmar
, but host immunity is likely to be relatively more robust given higher transmission intensity compared to western Thailand and Yunnan, where migrants from Myanmar account for the majority of infections
[19, 43–46]. Selective drug pressure also varies substantially due to differences in access to quality diagnosis and treatment. Populations included in the present study live in areas that government and international NGO health services largely fail to reach, and community based organizations are unique providers of malaria control interventions. The low prevalence of pfmdr1 amplification in Karen study sites may be due in part to these malaria programmes
 that for 10 years have provided directly observed mefloquine-artesunate that may modify selective pressure for pfmdr1 amplification
The pfcrt K76T mutation is a highly predictive marker of CQ resistance
. Replacement of K76T mutants with wild-type parasites following withdrawal of drug pressure has forecast the return of CQ clinical efficacy
[49, 50] and some authors have proposed tracking this mutation to identify populations in which CQ could be used as partner drugs in ACT
. The very low prevalence of wild-type pfcrt in the present study is consistent with data from nearby areas and suggests ongoing selective pressure for CQ resistance. Although decades have passed since CQ was officially recommended by Myanmar or any of its neighbours for the treatment of P. falciparum infection, CQ remains the treatment of choice for P. vivax and presumptive malaria. The use of RDTs capable of identifying both P. falciparum and P. vivax species may modestly diminish this selective pressure, but tracking the prevalence of wild-type pfcrt is likely to remain a low priority for resistance surveillance for the next several years. However, pfcrt could become an important genetic marker if the recommended treatment of P. vivax changes from CQ to ACT in response to the recent emergence of CQ-resistant P. vivax in western Thailand
Consistent with the a priori hypothesis, the prevalence of elevated pfmdr1 CN was higher among isolates obtained from active screening participants than from isolates obtained from febrile clinical patients. Resistance surveillance currently relies on samples obtained from clinically symptomatic malaria patients; containment efforts may need to more aggressively target subclinical infections that could serve as a reservoir for the spread of drug resistance if future studies validate an association between subclinical infection and resistance alleles. However, there are several reasons to approach this finding with caution. First, to the authors’ knowledge this is the first study to evaluate pfmdr1 CN in a predominantly asymptomatic population. Zhong et al. found no difference between symptomatic and asymptomatic study volunteers in Kenya in the frequency of pfmdr1 point mutations, or of the K76T mutation of pfcrt. Second, consistent with Zhong et al. the current study found no difference in K76T mutation prevalence between screening and clinical participants, though prevalence of wild-type pfcrt was low in each region and results may be confounded by ongoing CQ drug pressure as described above. Third, a robust empirically derived conceptual model is lacking to adequately capture the complex interactions between transmission intensity, host immunity, drug pressure, fitness cost, compensatory mutations, clonality of infection, de-amplification
[27, 55] and other factors expected to influence the relative prevalence of resistant and wild-type parasites in clinical and asymptomatic infections. Clonality of infection is likely to play an important role, as described above, but the prevalence of monoclonal infection in Myanmar is poorly quantified. A single study from central Myanmar documented mono-infection in 32% of clinical isolates
, but the relevance of these data to asymptomatic populations in border regions is unclear. The final reason to interpret with caution the association between subclinical infection and genetic resistance is the high variance of pfmdr1 CN estimates found among screening participants that led to the exclusion of 55% of samples. Future studies are needed to validate the accuracy of pfmdr1 CN estimates based on filter-paper blood samples collected from subclinical populations. Despite these caveats, findings presented here highlight the need to conduct additional studies on the contribution of subclinical infection to the epidemiology of drug-resistant malaria.
Although mosquitoes and migrating humans frequently carry parasites across international boundaries, it is not clear whether findings from neighbouring countries should be extrapolated to populations living in Myanmar. Individuals seeking care from border clinics by definition have access to diagnostic and treatment protocols and other resources of neighbouring countries that may differ substantially from those available in remote and conflict-affected areas of Myanmar. Additional studies are necessary to determine the appropriate geographic scale for monitoring the spread of drug resistance.
Systematic review of in vivo, in vitro and molecular resistance studies from Myanmar and its border regions identified relatively few data from populations living inside the country, particularly in Tier 1 and 2 areas of highest priority for resistance containment. The paucity of available information is consistent with output from the under-developed health research capacity in the country as a whole: Myanmar ranks 218th out of 224 countries in number of publications in medicine per capita (0.4 per 100,000 people)
. Armed conflict and large-scale population displacement are established causes of disruptions in health services and disease surveillance
, and can result in “stability bias”, defined as the “systematic under-sampling of populations and health threats in contexts of conflict and instability”
. The lack of overlap between previous resistance study locations and displaced populations in Myanmar documented in the present study suggests that stability-bias may contribute to within-country distribution of evidence available for resistance surveillance. The analysis presented here highlights areas with documented population displacement, but does not capture all conflict and human rights violations experienced by other communities. For example, approximately 92% of households in Chin State may experience forced labour
 and other human rights violations that have been associated with prevalent malaria infection
. The instability of sites participating in this study was tragically validated by violent events since data collection completed in 2009: one Karen site and all Kachin sites were displaced due to military attacks. Some health workers have continued to implement disease control interventions in relocated areas; and they remain willing to participate in resistance surveillance activities. One community-based organization is establishing a site to monitor parasite clearance time.
There were several important limitations to this study. Twenty P. falciparum isolates did not amplify either wild type or K76T alleles and it is not possible to exclude the presence of other rare resistance alleles such as SVMNK that were not assessed. None of these twenty isolates produced valid copy number estimates, and they do not modify our primary conclusions that pfmdr1 amplification may be more common among subclinical infections, and that that wild type pfcrt remains rare in these three regions of Myanmar.
The precision of pfmdr1 CN estimates determined from subclinical infections was low for 44% of isolates, and 16% of isolates had DNA concentrations below the lower limit of reliable copy number determination, even after four-fold DNA concentration with glycogen-acetate precipitation. As noted above, future studies are necessary to validate the use of filter-paper samples to estimate gene CN from low-parasitaemia isolates such as those collected from subclinical infections. This study was not designed to elucidate the mechanisms or relative contributions of factors influencing the prevalence of resistant parasites in subclinical populations. For example, clonality of infection was not determined and parasite DNA from prior infections was unavailable to distinguish re-infection from recrudescence.
The cross-sectional design limits causal inference, and findings related to measures of association should be considered preliminary, as noted above. Study participants lived in areas where quality malaria control services were ongoing for three to 8 years, and findings may not apply to intervention-naïve populations. Samples were collected in 2008–2009 and may not reflect the prevalence of resistance markers in 2012. For example, areas of western Thailand north of Maesot, immediately across the border from one of the present study sites, experienced a sharp rise in the proportion of slow-clearing parasites between 2008–10
The number of P. falciparum-positive isolates (290) was lower than anticipated by the study design (540) despite screening 4,591 villagers and 988 febrile clinical patients. Logistical constraints delayed data collection in Karen and Kachin areas until the lower-transmission, dry-season months of February and March, with extended clinical sampling through June. The study was conducted in areas of active malaria control, and the success of these programmes likely contributed to the low prevalence of P. falciparum and low RDT positivity rates. The smaller than anticipated number of P. falciparum positive samples limited the precision of estimates of genotype prevalence in subclinical and clinical infections, and poor statistical power precluded comparisons across geographic regions. Nevertheless, the number of isolates available for genotyping compares favourably to studies conducted in more stable areas, and data from individuals with subclinical P. falciparum infections is unique among published studies from this region.